Nonuniform oil shale rubble beds were retorted using the 10-ton retort at the Western Research Institute. In a series of 10 tests, two different size distributions of oil shale rubble were loaded into separate portions of the retort cross section. This loading procedure provided permeability ratios in the range of 0.019 to 0.97 between the two shale size distributions, and the fractional cross-sectional area of the smaller region ranged from 0.05 to 0.50 of the total retort cross section. In each retorting test, the gas flow distribution was measured at cold flow conditions, and the nonuniform rubble bed was retorted using direct combustion with air or air-and-steam mixtures. The first nine retorting tests demonstrate that the oil yield from the less permeable region decreases in proportion to the square root of the permeability ratio between the two shale size distributions. The measured oil yields correspond to the combined predictions of one-dimensional model calculations for each portion of the rubble bed. The dimensionless response times of gas tracers also correlate with the measured oil yields. A higher oil yield and retort sweep efficiency are demonstrated by burning propane at the top of the more permeable region in the tenth nonuniform retorting test. The propane combustion increases the gas pressure drop in the more permeable region, so more of the injected air is distributed to the less permeable region. For a severe permeability ratio of about 0.02, propane combustion improves retorting sweep efficiency from 50% to 80% of the rubble bed and increases oil yield from 16.5% to 32.2% of Fischer assay. The corresponding incremental energy production as oil and gas is 2.4 times greater than the cumulative energy content of the injected propane. 18 refs., 7 figs., 1 tab.